Вы находитесь на странице: 1из 8

Nonlinear dynamics of multi-disc rotor in dry friction bearings

Žigulić R a, Butković M a,b,c and Braut S a

Faculty of Engineering, University of Rijeka, Vukovarska 58, 51000 Rijeka, Croatia, Europe
Polytechnic of Karlovac, Ivana Meštrovića 10, 47000 Karlovac, Croatia, Europe
ALSTOM, Mala Švarča 155, 47000 Karlovac, Croatia, Europe

In this paper the FEM is used for calculating the transient response of horizontal multi-disc rotor
supported by two dry friction bearings. The nonlinearity in numerical simulation is introduced on the
basis of contact force theory through the deformation vector at contact point. The nonlinear system of
differential equations is solved by using modificated Newmark integration scheme associated with the
HHT α method. The results of calculations are compared for bearings with different radial clearances.
Detailed results are given in the form of diagrams of displacement amplitude in horizontal and vertical
direction as well as in the forms of whirl orbits. Also the diagrams of contact angle and normal contact
force in the contact point are given. Results of numerical model are compared with experimental results
obtained on the test rig during the rotor′s run down.

1. Introduction

Nonlinear dynamics of unbalanced rotor in the dry friction bearings, from the many points of view is very similar to
the other rotor-stator contact problems like blade-seal rub events, motion of the rotor in auxiliary bearings during
unexpected fault of magnetic bearings etc. In this case rotor has large unbalance forces and during start up and run
down passes through the range of critical speeds which can cause a total destruction of rotor and bearings. Significant
rotor vibrational amplitude increase in the area of critical speed is caused by sudden increase of contact force between
rotor and stator in that area. Thereby at horizontal rotors, which are by their nature (because of their weight) in constant
contact with bearings, taking and loosing of contacts are possible and they are connected with impacts and noise.
The most of theoretical and experimental researches, in this area, up to now are based on the Jeffcott rotor with the
retainer bearing. These investigations show that purely harmonic motion during the contact exist, although the complete
system is nonlinear [1]. The influence of vibrational parameters such as internal damping of the rotor, rotor-bearing
friction coefficient, contact stiffness, mass, stiffness and damping of the bearing, on the vibrations of the rotor as well as
on the transient vibrations during start up and run down are described in [2]. Experimentally verified expressions for
determining of normal contact force, also for determining of friction coefficient for noncircle shapes of dry friction
bearings are given in [3]. Paper [4] deals with the analysis of the influence of support damping on the disk
displacements as well as on the contact force between journal and back-up bearing’s inner side of active magnetic
bearing (AMB) system. Also, useful values of contact parameters as well as journal damping are given. Mentioned
papers, are dealing with rolling of the rotor on a vibrating stator with slip. It’s possible, in special cases, during the
passage through the critical speeds under special terms, that diverging backward precession, characterized with rolling
without slip of the rotor on vibrating stator, will appear [5], [6]. Paper [6] also confirms the identity of numerical results
obtained on the basis of collision and contact force theory. Multiple periodic solutions, quasi-periodic solutions and
chaotic responses, for selection of specific back-up bearing parameters of Jeffcott rotor in two auxiliary bearings, are
analyzed in [7].
There are only a few papers dealing with modeling of rubbing of a real rotors and applying the results from the
Jeffcott rotor to the more complex rotordynamical systems. Dry friction bearing clearance influence on steady-state
dynamic behavior of complex flexible rotor modeled by finite elements is shown in [8]. Paper [9] deals with the
synchronous response of continuous rotor, in the dry friction bearing, excited by unbalance. Experimental verification
of the analytical model of more complex rotor in one auxiliary bearing with clearances, based on the finite elements
method, is given in [10]
In present paper the FEM is used, and transient response of the three-disk rotor in dry friction bearings is calculated
using the direct time-integration method.
2. Results of Measurements

Test rotor has three disks and was tested in adapted balancing machine with rigid brass bearings. The diameter of the
journal is 15 mm. Bearings have radial clearance s=50 10-6 m. Bearings were oiled only by few drops of oil (µ=0.05).
Driving unit was tiristor controlled electromotor which variates rotor speed from n=0 rpm to n=3800 rpm.
For rotor vibration measurements Schenck’s Vibroport 41 was used with non-contact displacement pickups
VIBRONECS IN-085, fig. 1. The sensitivity of pickups were 8 mVµm-1. Various vibration analysis were performed.
Natural frequencies are determined by means of passage through the critical speeds (so called single tracking method),
because in case of using single shot method (when rotor is hit with impact hammer) an very large force for creating
disturbance on rotating rotor is
-Y necessary. Fig. 2. and fig. 3. show
Measuring run down measurements of first
sensor 2 harmonic of shaft vibration
displacement in horizontal and
Reference Rigid Rigid Rigid
vertical direction at disk No.3.
Driving sensor disk 1 disk 2 disk 3
Because of nonlinearities, it’s
unit Bearing 1 Bearing 2
not simple to obtain exact critical
speed values from amplitude
-Z diagrams, and therefore they are
determined from belonging phase
angle diagrams.
Fig 4. shows cascade diagrams
Flexible -X Measuring of disc No.3 vibrations in
coupling sensor 1 horizontal direction during run
down of rotor. Spectrum shoots
Figure 1. Scheme of the test rig with measuring equipment show that unbalance first harmonic
influence is dominant in relation to other
subharmonic and superharmonic components, and
that rotor internal damping and weight influence
is negligible.
Diagram with similar frequency contents is
obtained also in vertical direction. Fig 5. and fig. 6.
show orbits at disc No. 3 at rotor horizontal and
vertical critical speeds n=1880 rpm and n=2024 rpm.

3. Numerical Analysis

The equilibrium equation of the nonlinear

rotordynamic system can be written:

[M ]{q&&} + { f (q, q& )} = {g (q, t )} (1)

Figure 2. Run down, single tracking vs. speed, horizontal direction
with defined initial conditions {q0 } and {q&0 } .
Here, {q} is vector of the generalized coordinates,
[M] is a inertia matrix, { f (q, q& )} is a vector of the
internal forces containing the elastic, reactive,
dissipative and gyroscopic forces while
{g (q, t )} is a vector of the external forces due to
unbalance and weight. Regarding the use of FEM
in paper, equations of motion (1) are obtained by
assembling of equations of motion for the finite
elements, by means of which, rotordynamic
system is modeled. 3-D rotordynamic finite
element [11], based on theory of Bernoulli-Euler
or Timoshenko beam, is used for shaft modeling.
This beam model, defined with four degree’s of
freedom for each node enables the usage of
numerical damping in calculation. Hereafter a
Figure 3. Run down, single tracking vs. speed, vertical direction finite rigid disk element [11], also defined with
Rotational speed Ω Hz

Critical speed ω Hz
Figure 4. Cascade diagram, disc. No 3., horizontal direction

Vertical displ. Y m m


600 100

- 100
- 200
- 300
200 - 750 - 500 - 250 0 250 500 750
Vertical displ. Y m m

Horizontal displ. X mm

Figure 5. Rotor orbit at disc No. 3, n=1880 rpm


- 200

- 400

- 600
- 200 - 100 0 100 200
Horizontal displ. X mm

Figure 6. Rotor orbit at disc N. 3, n=2024 rpm Figure 7. Dry friction bearing finite element
four degrees of freedom in each node, is used in the calculation. Defining a dry friction bearing as a finite
element represents a task of the paper. Its equations of motion, obtained by using of Hamilton principle, are

[m ]{q&& }+ ([c ]− Ω [m ]){q& }+ [k ]{q }= {f }+ {f }

e e e e
e e e
e (2)

thereby dry friction bearing finite element in this case has six degrees of freedom q e = {u1 v1 α1 β1 u 2 v2 } T. Vector { }
{f } refers to unbalance linear external forces while vector {f } refers to nonlinear external forces, in accordance to
e e
contact force theory:

{f }= {− N cos γ + T sin γ
− N sin γ + T cos γ 00 N cos γ − T sin γ N sin γ + T cos γ }T (3)

In the previous expression N is normal contact force, T=µN is a friction force while γ is contact angle. In the expression
(2) inertia matrix [me], damping matrix [ce] and stiffness matrix [ke] are diagonal (only diagonal elements are specified):

[m ] = {m m
1 1 J a1 J a1 m2 m2 }T ; [c ] = {0 0 0 0 c
2x c2 y }T ; [k ] = {0 0 0 0 k
2x k2 y }T . (4)

while gyroscopic matrix is of unsymmetric type:

0 0 0 0 0 0
0 0 0 0 0 0

[m ] e
0 0
0 − J p1
J p1 0 0
 (5)
0 0 0 0
0 0 0 0 0 0
 
0 0 0 0 0 0

The nonlinearity in numerical simulation is introduced through the deformation vector at contact point {δ } . Thereby
nonlinear rubbing model, based up on contact force theory, is used. Normal contact force in contact point, in the case of
transient response calculation is, fig.7:

{N } = K Cδ e jγ if δ > 0 (6)

Magnitude of contact point deformation vector {δ} is equal to:

δ = (u1 − u 2 )cos γ + (v1 − v2 )sin γ − s (7)

where s is radial clearance of the rotor. Magnitude δ is a function of contact angle γ, calculated from the actual location
of contact:

v1 − v2 (8)
γ = Arc tan
u1 − u2

The transient response of the system (1) is obtained by using modificated Newmark integration scheme (HHT α
method was used to solve instability problem in the case of vibrational systems with damping [13], especially if contact
damping in the equation (6) is used). For the solving of a system of nonlinear equations

( )
r {q}t + ∆t = [M ]{q&&}t + ∆t + { f }t + ∆t − {g}t + ∆t = 0 (9)

rit + ∆t
where r is residium, a Newton – Raphson method was used. Thereby a residue norm criterion ≤ ε r is used for
r1t + ∆t

convergence assurance where residue norm is equal to rit + ∆t = {r ∆ } {r ∆ } .

t+ t T t+ t
3.1. Numerical example
Rotor shown on fig. 1. is modeled as overhanging beam and is discretized by 7 finite elements and 8 nodes. Rotor is
made of steel material (Young’s modulus E=2.1⋅1011 Nm-2 and mass density ρ=7800 kgm-3.The area on the left from the
first bearing represents the first finite element with the length l1=41.7 mm and diameter d1=15 mm. The second finite
element has the same diameter, while its length is l2= 33 mm. On the left-hand side of the rigid disk 1 the third finite
element is situated with length l3=28 mm and diameter d3=9.3 mm. Between the rigid disks 1 and 2, and 2 and 3, there
are finite elements 4 and 5 with the same diameters as finite element 3. The belonging lengths are l4=113.8 mm and
l5=114.5 mm. On the right-hand side of the rigid disk 3 the finite element 6, with the same diameter as finite element
3,4 and 5 and length l6=34 mm, and finite element 7 with the length l7=35 mm and diameter d7=15 mm, are situated.
Coefficients of the shaft internal damping are ηV=10-3, ηH=0.
The discretized model has four concentrated masses. The flexible coupling is modeled as a mass m1=0.1 kg and
principal mass moments of inertia Jx1=10-5 kgm2 with Jy1=Jz1=5.5 10-6kgm2. Three rigid disks are modeled as masses
m2=m3=m4=1.55 kg and principal mass moments of inertia Jx2=Jx3=Jx4=3⋅10-3kgm2 and
Jy2=Jz2=Jy3=Jz3= Jy4=Jz4=1.6 10-3kgm2. Eccentricity of a mass unbalance on the disk No 2 is a=4 10-5 m.
Dry friction bearings have mass ml=0.2 kg, contact stiffness is KC=2 108 Nm-1 and dry friction coefficient in the
contact point is µ=0.05. Dry friction bearings are mounted on the springs with horizontal kx=5 105 Nm-1and vertical
ky=4.5 108 Nm-1 stiffnesses as well as on the dampers with horizontal cx=20 Nsm-1and vertical cy=2000 Nsm-1dampings.

Figure 8. Numerical amplitudes of horizontal (left) and vertical (right) displacement, s1=50 10-6 m

Figure 9. Numerical amplitudes of horizontal (left) and vertical (right) displacement, s2=100 10-6 m

Figure 10. Numerical amplitudes of horizontal (left) and vertical (right) displacement, s3=200 10-6 m
Figure11. Nonlinear response on disk No3 at first horizontal (left) and vertical(right) natural frequency

Figure 12. Nonlinear response at bearing 2 (rotor part) at first horizontal (left) and vertical(right) natural frequency

Figure 13. Nonlinear response at bearing 2 (stator part) at first horizontal (left) and vertical(right) natural frequency

Figure 14. a) Figure 14. b)

Figure 14. a) Deformation δ in frequency range
0-63 Hz
b) Deformation δ in frequency range
28.16-37.14 Hz
c) Deformation δ in frequency range
25.16-38.14 Hz

Figure 14. c)

Figure 15. a) Contact angle γ in frequency range Figure 15. b) Contact angle γ in frequency range
0-63 Hz 25.16-38.14 Hz
Used parameters of numerical method (HHT) are α=0,08 and residue norm criterion εr=10-6.
The most essential results of the calculation of rotor’s nonlinear transient vibration are shown on figures 8....15.
Results of rotor’s start up wasn’t showed because greater vibration response would be obtained according to theory, and
experiment would cause a test rig and measuring equipment problems. Also, same observations was registered on a test
rig, where there was problems with leaving the instability when rotor was running into the area of first horizontal and
vertical natural frequency. Results of rotor’s run-down calculation (Ω=400-10t rads-1) for three radial clearances values
s1=50 10-6 m, s2=100 10-6 m i s3=200 10-6 m are shown on figures 9…11. Duration of run-down was t=40 s.
Characteristic transient response curve bending and expansion of a critical speed area (in relation with linear
calculation) was obtained because nonlinear model was used. Decreasing of natural frequency with increasing of radial
clearance s was observed.
More details about a rotor’s behavior in the dry friction bearings, refer to the rotor with radial clearance s1=50 10-6
m are shown on figures 11…15. Rotor orbits calculated for the disk No.3, caused by unbalance excitation, at first
horizontal(29.4 Hz) and vertical(32.5 Hz) natural frequencies are shown on figure 11. Each orbit consists of 30
sequential calculated points. Orbits of the bearing No. 2 at first horizontal and vertical natural frequencies, as an
example of motion of the dry friction bearing finite element, are shown on figures 12. (rotor) and 13. (stator).
Displacements of the stator part are expected to be smaller than displacements of rotor part and in both cases (horizontal
and vertical resonance) represent almost horizontal curves because stator part of the dry friction bearing has
considerably higher stiffness in vertical direction related to horizontal stiffness. Orbits of rotor part (fig 12.) show that
rotor near the first horizontal natural frequency is in permanent contact with stator part but only on the lower part of the
bearing while near the first vertical natural frequency rotor is also in permanent contact with stator, rolling with slip
around the stator on the boundaries of radial clearance, fig 12. (right). Above mentioned statements could be better
explained on the figure 14. showing deformation magnitude δ during rotor shut-down diagrams as well on the figure 15.
showing the contact angle γ. On the basis of these diagrams it is possible to conclude that rotor is, almost all the run-
down time period, in the contact with lower part of the dry friction bearing due its weight. In the areas of first horizontal
and first vertical natural frequency rotor is also in contact with bearing, with relatively high magnitudes of deformation
vector {δ} and by this with high magnitudes of normal component of contact force N. The difference in the motion near
the first horizontal and first vertical natural frequency could be seen from the diagrams of calculated contact angles γ,
figure 15. In the area of first horizontal natural frequency this angle slowly variates about the value γ=2700 which means
that rotor, as was earlier mentioned, is moving only in the lower part of the bearing, while in the area of first vertical
natural frequency angle γ assumes values from 0 to 3600. From the fig. 14. can be seen that rotor bounces between first
horizontal and first vertical as well as above the first vertical natural frequency. These bounces originate from the
looseness of contact (δ<0). At these bounces the contact angle also has different values from 0 to 3600. Mentioned
behavior can be explained by motion in the vicinity of vertical natural frequency but still quite far that during the
bouncing the contact is restored with some point on the upper part of bearing. Higher number of bouncing above the
first vertical natural frequency is related to the bouncing between first natural frequencies can be explained by higher
excitation force which is rising the rotor from the contact in the lower part of bearing.
Similar behavior of rotor can be seen at clearances s2 and s3 but with some differences. At clearances s2=100 10-6 m
rotor also rolls with slip around the bearing on the boundaries of radial clearance but the frequency range of this motion
is narrower. Also with this clearance disappears bouncing between two natural frequencies. At clearances s3=200 10-6 m
rolling with slip near vertical critical speed disappears and this phenomena is substituted by bounces at the area of first
vertical critical speed.

4. Conclusions

In the paper problem of transient nonlinear vibrations of the rotor in two dry friction bearings is elaborated
theoretically, numerically and experimentally. Forward whirl in connection with rolling with slip of the rotor on the
vibrating stator is investigated. Natural frequencies obtained numerically and experimentally show very good
agreement. Also response curves of the rotor during rotor run-down obtained by measurements and by calculations
show very good agreement. Because of high usage time of CPU (3.5 hours for one example on PENTIUM III at 733
MHz with 256 Mb RAM), in the case of nonlinear calculations low degree of discretization of numerical model (7 finite
elements) is used. That discretization still gives the satisfactory results for the explanation of some nonlinear
phenomena. Combination of numerical and experimental approach has advantage that it is not necessary to measure the
vibrational characteristics in all rotor′s positions, especially in the inaccessible positions (bearings with dry friction) but
only on some of rotor′s discs. Additionally, numerical results give whole «history« of rubbing events in the dry friction
bearings during the rotor′s start up and run down.
Presenting nonlinear rubbing model can be used in the solving of three different problems in industrial praxis. First,
model can be applied for rotors in magnetic bearings in the case that AMB system fails due to dissolution of electric
supply or other malfunctions. Also model can be used in solving of classical rotors in hydrodynamics journal bearings
during the contacts due to insufficient oil pressure. Finally simultaneous rubbing on several contact points between rotor
and stator (journal-bearings, blade-seals) can be modeled. That problem will be matter of our further investigations.

5. References

[1] Edbauer R.,Meinke P., Muller P.C., Wauer J.: Passive Durchlaufhilfen beim Durchfahren biegerkritischer
Drehzahlen elastischer Rotoren, VDI-Berichte Nr.456,1982, pp. 157-166.
[2] Markert R., Wegener G.: Dynamik von elastischen Rotoren in Fanglagern, Schwingungen in Rotierenden
Maschinen III, Universitaet Kaiserslautern, 1995., pp. 20-30.
[3] Simon, U., Brommund, E.: Periodische Bewegungen einer Pendelzentrifuge in einem mehreckigen Fanlager,
Schwingungen in rotierden Maschinen IV, Viewig, pp.181-188.
[4] Ishii T., Kirk R. G.: Transient Response Technique Applied to Active Magnetic Bearing Machinery During Rotor
Drop, ASME Journal of Vibration and Acoustics, 1996., pp. 154-163.
[5] Yanabe, S., Nishimura, T., Kaneko, S.: Backward Whirl of Uniform Vertical Rotating Shaft Supported by
Unlubricated Sleeve Bearing, Proceedings of the 9. IFTOMM World Congress, Milano 1995., Vol. 2. pp. 1253-1256.
[6] Yanabe, S., Kaneko, S., Fukushima, T., Kanemitsu, Y., Tomi, N., Sugiyama, K.: Rotor Vibration due to Collision
with Annular Guard During Passage Through Critical Speed, Transaction JSME, Vol.60, No. 571, 1994, pp 21-28.
[7] Wang, X., Noah, S.: Nonlinear Dynamics of a Magnetically Supported Rotor on Safety Auxiliary Bearings, ASME
Journal of Vibration and Acoustics, 1998., Vol. 120, pp. 596-606.
[8] Xie H., Flowers G.T. Feng L., Lawrence C.: Steady-State Dynamic Behavior of a Flexible Rotor With Auxiliary
Support from a Clearance Bearing, Journal of Vibration and Acoustics, 1999., Vol. 121, pp. 78-83.
[9] Wegener G., Markert R., Pothmann K.: Steady-State-Analysis of a Multi-Disk or Continuous Rotor with one
Retainer Bearing, Proceedings of the 5th International Conference on Rotor Dynamics, Darmstadt, September 7-10,
1998., pp. 816-828.
[10] Lawen J. L. Jr., Flowers G.T.: Synchronous Dynamics of a Coupled Shaft/Bearing/Housing System with Auxiliary
Support from a Clearance Bearing, Journal of Engineering for Gas Turbines and Power, 1997., Vol. 119, pp. 430-435.
[11] Nelson, D., McVaugh, J.M.; The Dynamics of Rotor-Bearing Systems Using Finite Elements, Journal of
Engineering for Industry, May 1976, pp. 593.- 600., Paper No. 75-WA/DE-19.
[12] Žigulić, R.: Dynamical Analysis of Rotor, Doctoral Thesis, Technical Faculty University of Rijeka, 2001 (in
[13] Geradin, M., Kill, N.: Nonlinear Dynamic Analysis of Flexible Rotor, 4th International Conference on Vibrations in
Rotating Machinery, Edinburgh, 1988., Report VA-50, pp. 1-9.